Research map / 3D vision / Human pose estimation
Key Frame Proposal Network for Efficient Pose Estimation in Videos
Yuexi Zhang, Yin Wang, Octavia Camps, Mario Sznaier
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use learned K-FPN key frames instead of uniform sampling (ablation) | K-FPN + HPIM · Penn Action | PCK: 79.3 -> 98.0 | Improved |
| use learned K-FPN key frames instead of best of 100 random subsets (ablation) | K-FPN + HPIM · Penn Action | PCK: 96.4 -> 98.0 | Improved |
| deeper backbone ResNet34 to ResNet101 (ablation) | K-FPN · Penn Action validation | PCK avg: 97.1 -> 97.7 | Improved |
| deeper backbone ResNet18 to ResNet50 (ablation) | K-FPN · sub-JHMDB validation | PCK avg: 97.3 -> 97.9 | Improved |
| online instead of batch key frame selection (Tb=30, To=10) (ablation) | K-FPN · sub-JHMDB | PCK mean: 94.5 -> 94.4 | No clear effect |
| online instead of batch selection (ablation) | K-FPN · sub-JHMDB | avg key frames: 16.3 -> 15.2 | Improved |
| use K-FPN selected frames under illumination changes (60% perturbed) | K-FPN vs baseline · sub-JHMDB | PCK: 90.2 -> 92.1 | Improved |
| use K-FPN selected frames under blurring (60% perturbed) | K-FPN vs baseline · sub-JHMDB | PCK: 87.9 -> 91.0 | Improved |
| use K-FPN selected frames under occlusions (60% perturbed) | K-FPN vs baseline · sub-JHMDB | PCK: 85.4 -> 90.4 | Improved |
| replace frame-based baseline with K-FPN ResNet34 | K-FPN vs baseline [cite] · Penn Action | running time ms: 11.3 -> 5.3 | Improved |
| replace frame-based baseline with K-FPN ResNet18 | K-FPN vs baseline [cite] · sub-JHMDB | running time ms: 10.0 -> 4.7 | Improved |
Closest papers by meaning
- Deep Dual Consecutive Network for Human Pose Estimation (CVPR 2021 · similarity 0.89)
- Temporal Feature Alignment and Mutual Information Maximization for Video-Based Human Pose Estimation (CVPR 2022 · similarity 0.88)
- Detect-and-Track: Efficient Pose Estimation in Videos (CVPR 2018 · similarity 0.87)
- Learning Temporal Pose Estimation from Sparsely-Labeled Videos (NeurIPS 2019 · similarity 0.87)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.86)
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
